| Resumo: | The fast growth of Internet-connected embedded devices raises new challenges for the traditional network design, such as scalability, diversity, and complexity. To endorse these challenges, this thesis suggests the aggregation of several emerging technologies: software-defined networking (SDN), fog computing, containerization and sensor virtualization. This thesis proposes, designs, implements and evaluates a new solution based on the emergent paradigm of SDN to efficiently manage virtualized resources located at the network edge in scenarios involving embedded sensor devices. The sensor virtualization through the containers provides agility, flexibility and abstraction for the data processing, being possible to summarize the huge amount of data produced by sensor devices. The proposed architecture uses a software-defined system, managed by a Ryu SDN controller, and a websocket broker written from scratch that analyses the messages sent to the controller and activates containers when required. Performance and functional tests were performed to assess the time required from activating the sensor containers to being able to communicate with them. The results were obtained by sending four ICMP packets. The best time response results were obtained by the proactive controller behavior mode, when compared to the hybrid and reactive modes. This thesis contributed to fill the gaps in the area of IoT or sensor networks, concerning the design and implementation of an architecture that performed on-demand activation of offline IoT fog computing resources by using an SDN controller and sensor virtualization through containers.
|
|---|